An express package informationization monitoring method and system based on an internet of things

Abstract

The application provides an express package informationization monitoring method and system based on the Internet of Things, which realizes real-time, whole-process and intelligent monitoring and early warning of express packages by combining edge computing and deep learning technology. Specifically, sensors and miniature cameras are installed on the express package to collect package information and environmental data. The edge computing device is responsible for real-time data collection and preliminary processing nearby to optimize the use of computing resources and improve data processing speed. The application also introduces a zero-shot learning model to predict new and previously unseen situations based on the rules learned from training samples. In addition, the application trains a deep learning model and monitors the state of express packages in real time to immediately warn of potential problems. This method greatly improves the efficiency, accuracy and real-time performance of package monitoring, ensuring the accuracy and efficiency of express operations.

Description

Technical Field

[0001] This invention belongs to the field of express logistics, and more specifically relates to an information monitoring method and system for express packaging based on the Internet of Things. Background Technology

[0002] In the past, parcel monitoring relied primarily on manual inspection, with couriers periodically checking parcel status. This method was inefficient, error-prone, and limited to fixed time points, failing to achieve end-to-end monitoring. With rapid societal development and the booming express delivery industry, this labor-intensive parcel management approach can no longer meet the industry's demands for efficiency and accuracy.

[0003] The trend of modern technological development is towards informatization, intelligentization, and networking. With the development of Internet of Things (IoT) technology, the widespread use of various sensors, and the gradual maturation of edge computing and cloud computing technologies, these technologies have brought new opportunities to the express delivery industry.

[0004] The Internet of Things (IoT) technology, through the deployment of various sensors and miniature cameras, can track and collect package status information and environmental data in real time. Edge computing technology can process the large amounts of collected data quickly and in real time locally, reducing the computing tasks on the cloud and improving overall data processing efficiency.

[0005] Furthermore, the rapid development of deep learning technology in recent years, especially the introduction of zero-shot learning methods, allows the model to learn and generalize patterns learned from training samples to seemingly unrelated datasets. This enables the model to predict new situations that did not appear in the original training samples. This provides a new solution for dynamic, complex, and ever-changing real-time monitoring tasks such as express parcel delivery.

[0006] Therefore, how to effectively combine cutting-edge technologies such as the Internet of Things, edge computing, and deep learning to build a highly efficient, accurate, and real-time information monitoring system for express parcels, in order to meet the growing monitoring needs, improve the accuracy and efficiency of express delivery operations, and ensure the safety of customers' parcels, has become an important research direction for the modern express delivery industry. Summary of the Invention

[0007] This invention addresses the shortcomings of traditional express parcel monitoring methods in terms of efficiency, accuracy, and real-time performance. By utilizing IoT technology, edge computing, and deep learning, it enables real-time, end-to-end, and intelligent monitoring of express parcels, providing accurate early warnings of potential risks and improving the efficiency and security of express delivery services.

[0008] To achieve the above objectives, the present invention employs the following technical solution: the method comprises:

[0009] Different sensors and miniature cameras are installed on express delivery packaging to collect package information and environmental data;

[0010] Edge computing devices collect and process data in real time nearby, reducing the computing load on the cloud and improving data processing speed;

[0011] Building a deep learning model: Build a zero-shot learning model, which learns and generalizes the patterns learned from training samples to seemingly unrelated datasets;

[0012] Training the deep learning model: The zero-shot learning algorithm is used to train the model on the collected data, so that the model can predict new situations that have not appeared in the original training samples;

[0013] Status monitoring and early warning: The system uses a trained model to monitor the status of express parcels in real time. If the model predicts a problem, it will immediately issue an early warning, such as warning that the parcel has been subjected to vibration or excessive temperature.

[0014] In one embodiment, the environmental data includes location, temperature, humidity, vibration, and light intensity.

[0015] In one approach, the data collection and preliminary processing include: data filtering and noise reduction, location data analysis, preliminary image analysis, and vibration data analysis.

[0016] In one approach, the zero-shot learning model includes:

[0017] The model architecture includes the following components:

[0018] Semantic embedding space: embeds semantic information of categories or states into a high-dimensional vector space, using word vector models;

[0019] a c =Embedding(Semantic_Info) c )

[0020] Among them, a c It is the semantic embedding vector of category c.

[0021] Visual embedding space: The fused feature vector z is projected onto the semantic embedding space to enable cross-space matching.

[0022] v = W v z+b v

[0023] Among them, W v and b v These are the weight matrix and bias vector for the linear transformation, respectively, and v is the visual embedding vector;

[0024] Compatibility function: Define a compatibility function F(v, a) c This is used to measure the similarity between visual embeddings and semantic embeddings; the compatibility function is as follows:

[0025] F(v,a c ) = v · Wa c

[0026] Where W is the weight matrix for learning, used to capture the relationship between the two.

[0027] Loss function design: To train the model and maximize the compatibility score of the correct class, a contrastive loss is used.

[0028]

[0029] Where γ is the marginal hyperparameter, c′ is the negative sample class, and the score of the correct class c is guaranteed to be at least γ higher than the score of the negative sample c′.

[0030] In one approach, training the deep learning model includes employing marginal contrastive loss to ensure that the matching score of the correct class with the sample is at least higher than the score of the incorrect class by a marginal γ.

[0031] Marginal contrast loss function:

[0032]

[0033] Where γ is the marginal hyperparameter, c′ is the negative sample class, and the score of the correct class c is at least γ higher than the score of the negative sample c′; F(v x ,a c The score represents the compatibility score between visual embedding and semantic embedding.

[0034] Model training process

[0035] Forward Propagation: Input Data: The fused feature vector z of each sample in the training set is input into the visual embedding network to generate a visual embedding vector v.

[0036] v = W v z+b v

[0037] Semantic embedding: using a predefined semantic embedding vector a c Represent each category;

[0038] Compatibility score: Calculate the compatibility score F(v,a) for each pair of visual embeddings and semantic embeddings. c );

[0039] Loss Calculation: Based on the marginal contrast loss function, calculate the loss for each sample and accumulate the total loss for the entire training set.

[0040]

[0041] Backpropagation and parameter update:

[0042] The gradient of the loss function with respect to the model parameters is calculated using the backpropagation algorithm, and the parameters θ are updated using the optimization algorithm (Adam):

[0043]

[0044] Where η is the learning rate. This represents the gradient of the loss function with respect to the parameters.

[0045] Regularization and overfitting prevention: Introducing an L2 regularization term to prevent the model from overfitting the training data: Where λ is the regularization intensity coefficient.

[0046] In one approach, the status monitoring and early warning system incorporates a multi-level risk assessment model to classify and process anomalies of different types and degrees, thereby accurately identifying potential risks.

[0047]

[0048] Where K is the number of risk categories, m k w represents the number of risk indicators of type k. k,i and S k,i (t) represents the weight and score, respectively.

[0049] On another front, an IoT-based information monitoring system for express packaging is provided. This system is applicable to the method described above. The system includes sensors and miniature cameras installed on the express packaging to collect package information and environmental data; a data collection device with edge computing capabilities for real-time data collection and preliminary processing; a zero-shot learning model for learning and generalizing from seemingly unrelated datasets based on patterns learned from training samples; a deep learning model training device that uses a zero-shot learning algorithm to train the collected data; and an early warning device for real-time monitoring of the express package status using the trained model, providing a real-time warning when the model predicts a problem.

[0050] Beneficial effects of this invention:

[0051] By leveraging IoT, edge computing, and deep learning technologies, this invention can significantly improve the efficiency of express packaging monitoring. Real-time monitoring, automated processing, and intelligent early warning can avoid excessive manual intervention and increase the speed of workflow operation.

[0052] Training and applying deep learning models can improve the accuracy of predicting the status of express parcels. Furthermore, the use of miniature cameras and sensors helps obtain more objective and detailed real-time data, improving the accuracy of the analysis results.

[0053] The real-time monitoring and intelligent early warning functions of this invention can help detect potential risks early, improve the speed of handling abnormal situations, and effectively ensure the safety of express parcels.

[0054] By processing data on edge devices, the computing load on the cloud can be effectively reduced, computing tasks can be efficiently allocated, and the overall performance of the system can be improved. Attached Figure Description

[0055] Figure 1 This is a flowchart of the method of the present invention;

[0056] Figure 2 A flowchart for establishing a deep learning model for this invention;

[0057] Figure 3 This is a flowchart of the deep learning model training process of the present invention. Detailed Implementation

[0058] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0059] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. To facilitate understanding, the invention will now be described more fully with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to make the disclosure of the invention more thorough and complete.

[0060] like Figure 1 As shown, an IoT-based information monitoring method for express packaging includes:

[0061] S1. Sensor Configuration: Different sensors (such as position, temperature, humidity, vibration, light, etc.) and miniature cameras are installed on the express packaging. These devices will be responsible for collecting package information and environmental data.

[0062] Determine the environmental parameters to be monitored, such as location, temperature, humidity, vibration, and light intensity, as well as whether image capture functionality is required. Based on these requirements, selecting the appropriate sensor type and model is crucial. For perishable goods, prioritize high-precision temperature and humidity sensors; for valuable or fragile items, focus on configuring high-sensitivity vibration sensors and positioning modules.

[0063] S2. Edge computing devices: These devices collect and process data locally in real time, reducing the computing load on the cloud and improving data processing speed, such as performing preliminary analysis on images captured by cameras.

[0064] Edge computing devices play a crucial role, responsible for real-time data collection and preliminary processing locally, thereby reducing the computing load on the cloud and improving data processing speed. The following is a detailed process for implementing edge computing devices.

[0065] S201. Equipment Selection and Deployment

[0066] First, it's necessary to select suitable edge computing devices. These include high-performance microprocessors (such as ARM architecture processors), sufficient memory and storage space, and modules that support wireless communication (such as Wi-Fi, Bluetooth, LoRa, or NB-IoT). The devices should be characterized by low power consumption and high efficiency to meet the needs of long-term operation and different transportation environments.

[0067] S202, Data Collection and Preprocessing

[0068] Edge computing devices collect multi-dimensional data about packages in real time by connecting to various sensors and cameras. This data may include temperature, humidity, vibration, light intensity, and images. To improve data processing efficiency, the collected data first needs to be preprocessed, including data filtering, noise reduction, and compression.

[0069] (1) Data filtering and noise reduction:

[0070] For continuous sensor data, filters can be used to remove noise, and low-pass filters (such as moving average filters) can be used to smooth temperature and humidity data.

[0071]

[0072] Where y[n] is the filtered signal, x[n] is the original signal, and M is the size of the filtering window.

[0073] (2) Preliminary data processing and analysis

[0074] Edge computing devices need to perform preliminary analysis on the collected data to extract useful information and reduce the amount of data transmitted to the cloud.

[0075] Location data analysis: Location data collected by the GPS module can be used to estimate and predict location using a Kalman filter to improve positioning accuracy.

[0076]

[0077] in, Let P be the state vector, P be the error covariance matrix, K be the Kalman gain, and z be the state vector. k The values ​​are measured, and A, B, H, Q, and R are the system matrix and noise covariance matrix, respectively.

[0078] Preliminary image analysis: For images captured by the camera, edge devices can perform preliminary image processing and analysis, such as object detection and classification. Feature extraction is performed using pre-trained CNN models (such as MobileNet).

[0079] Vibration data analysis: For vibration sensor data, the time-domain signal can be converted into a frequency-domain signal using Fourier transform (FFT) to analyze the vibration frequency characteristics.

[0080]

[0081] Frequency domain analysis can detect abnormal vibration frequencies, thereby determining whether the package has been subjected to abnormal impact.

[0082] Edge computing devices make local decisions based on preliminary analysis results. If the temperature exceeds a set threshold or abnormal vibration is detected, the system can immediately issue an early warning signal, reducing reliance on cloud resources. Early warning rules can be implemented through logical judgments:

[0083] If T>T threshold or V>V threshold Trigger alert

[0084] Where T is temperature, V is vibration intensity, and T threshold and V threshold This is a preset threshold.

[0085] like Figure 2 As shown, S3, establish a deep learning model: establish a deep learning model architecture, namely a zero-shot learning model, which can learn and generalize to seemingly unrelated datasets by learning the rules learned from training samples.

[0086] The aim is to develop a zero-shot learning (ZSL) model. This model learns patterns from training samples and can generalize to seemingly unrelated datasets, thereby achieving accurate prediction and classification of novel or unseen package states. The following are the detailed implementation steps, along with the mathematical formulas and theoretical foundations involved.

[0087] S301. Dataset Preparation and Integration

[0088] Combining the first two steps (S1 and S2), the system has collected a large amount of multi-dimensional data from sensors and edge computing devices. This data includes environmental parameters (such as temperature, humidity, vibration, and light intensity), location data, and image information. This data is then organized and labeled to build the deep learning model.

[0089] Data preprocessing:

[0090] Data cleaning: Remove missing values, outliers, and noisy data to ensure data quality. For example, use statistical methods to detect and remove outlier data points that exceed 3 standard deviations.

[0091] Data standardization: Normalizing data of different dimensions to make them have the same scale, including min-max normalization and Z-score standardization.

[0092] Feature Extraction and Fusion: Sensor Data Feature Extraction: Extracting statistical features (such as mean and variance) and time series features (such as autocorrelation) from sensor data such as temperature and humidity.

[0093] Image data feature extraction: High-dimensional image feature vectors are extracted using pre-trained convolutional neural networks (such as ResNet and VGG).

[0094] Data fusion: This involves combining features from different sources to form a unified feature representation. It includes feature concatenation and attention mechanisms.

[0095] z = Concat(z) sensor ,z image )

[0096] Among them, z sensor and z image denoted as the feature vectors of the sensor and image data, respectively, and z is the fused feature vector.

[0097] S302, Zero-Shot Learning Model Architecture Design

[0098] Zero-shot learning aims to enable a model to recognize categories or states that are not present in the training set. The core of the S3 phase is designing an effective ZSL model architecture, which typically includes the following components:

[0099] Semantic embedding space: Embedding semantic information of categories or states (such as text descriptions and attribute vectors) into a high-dimensional vector space. This is achieved using word vector models (such as Word2Vec).

[0100] a c =Embedding(Semantic_Info) c )

[0101] Among them, a c It is the semantic embedding vector of category c.

[0102] Visual embedding space: The fused feature vector z is projected onto the semantic embedding space to enable cross-space matching.

[0103] v = W v z+b v

[0104] Among them, W v and b v are the weight matrix and bias vector of the linear transformation, respectively, and v is the visual embedding vector.

[0105] Compatibility function: Define a compatibility function F(v, a) c This function measures the similarity between visual embeddings and semantic embeddings. The compatibility function is as follows:

[0106] F(v,a c ) = v · Wa c

[0107] Where W is the weight matrix for learning, used to capture the relationship between the two.

[0108] Loss function design: To train the model and maximize the compatibility score of the correct class, a contrastive loss is used.

[0109]

[0110] Where γ is the marginal hyperparameter, c′ is the negative sample class, and the score of the correct class c is guaranteed to be at least γ higher than the score of the negative sample c′.

[0111] S303, Zero-Sample Inference and Applications

[0112] During the deployment phase, the model needs to infer new, unseen categories or states. Based on the compatibility function learned during training, the model calculates a compatibility score between the new sample and the semantic embeddings of all possible categories, selecting the category with the highest score as the prediction result.

[0113]

[0114] Among them, C unseen For the set of unseen categories, For the predicted category.

[0115] Deploy the trained deep learning model to the cloud or edge server to ensure it can process data streams from edge devices in real time. During deployment, the model's computational complexity and response time must be considered, and model compression and quantization should be performed if necessary.

[0116] W quant =round(W×2) q )

[0117] Where q is the number of quantization bits, which reduces the model's storage and computational requirements.

[0118] The process of building a zero-shot deep learning model (S3) closely integrates sensor configuration (S1) and real-time data processing from edge computing devices (S2). Through systematic data preparation, model architecture design, training optimization, and deployment, the ZSL model maintains high efficiency and accuracy in predicting new or unseen package conditions. Through continuous optimization and iteration, the model adapts to dynamically changing logistics environments, significantly improving the intelligence and reliability of the entire express packaging information monitoring system.

[0119] like Figure 3 As shown, S4, Training the deep learning model: The zero-shot learning algorithm is used to train the collected data so that the model can predict new situations that have not appeared in the original training samples.

[0120] This step is based on the Zero-Shot Learning (ZSL) model established in the S3 stage. It uses a specific training algorithm to train the collected multi-dimensional data to ensure that the model can not only accurately identify the categories or states that appear in the training set, but also effectively predict and classify new situations that have not appeared in the original training samples.

[0121] S401. Data Preparation and Partitioning: In stage S3, data preprocessing, feature extraction, and fusion have been completed. Stage S5 first requires partitioning this prepared data to facilitate effective training, validation, and testing.

[0122] Data partitioning:

[0123] Training Set: Contains a large number of labeled samples used for model parameter optimization. It typically comprises 70%-80% of the total data.

[0124] Validation set: Used to adjust model hyperparameters and prevent overfitting, it typically comprises 10%-15% of the total data.

[0125] Test Set: Used to evaluate the model's performance on unseen data, typically comprising 10%-15% of the total data.

[0126] S402. Zero-Shot Learning Algorithm Selection and Configuration

[0127] Zero-shot learning aims to enable models to handle categories or states that do not appear in the training set. In this system, a compatibility function-based embedding method is adopted. The core idea is to map visual features and semantic features to the same embedding space and evaluate the degree of matching between them through a compatibility function.

[0128] Compatibility functions:

[0129] F(v,a c ) = v · Wa c

[0130] Where v is the visual embedding vector, a c Let W be the semantic embedding vector for category c, and W be the learned weight matrix.

[0131] S403. Loss Function Design and Optimization Objectives

[0132] To train a ZSL model so that the compatibility score of the correct class is higher than that of the incorrect class, a contrastive loss is typically used. In this system, a marginal contrastive loss is used to ensure that the matching score between the correct class and the sample is at least one marginal γ higher than the score of the incorrect class.

[0133] Marginal contrast loss function:

[0134]

[0135] Where γ is the marginal hyperparameter, c′ is the negative sample class, and the score of the correct class c is guaranteed to be at least γ higher than the score of the negative sample c′.

[0136] F(v x ,a c The score represents the compatibility score between visual embedding and semantic embedding.

[0137] S404, Model Training Process

[0138] Forward propagation:

[0139] Input data: The fused feature vector z of each sample in the training set is input into the visual embedding network to generate a visual embedding vector v.

[0140] v = W v z+b v

[0141] Semantic embedding: using a predefined semantic embedding vector a c Each category is represented.

[0142] Compatibility score: Calculate the compatibility score F(v,a) for each pair of visual embeddings and semantic embeddings. c ).

[0143] Loss Calculation: Based on the marginal contrast loss function, calculate the loss for each sample and accumulate the total loss for the entire training set.

[0144]

[0145] Backpropagation and parameter update:

[0146] The gradient of the loss function with respect to the model parameters is calculated using the backpropagation algorithm, and the parameters θ are updated using the optimization algorithm (Adam):

[0147]

[0148] Where η is the learning rate. This represents the gradient of the loss function with respect to the parameters.

[0149] Regularization and overfitting prevention: Introducing an L2 regularization term to prevent the model from overfitting the training data:

[0150]

[0151] Where λ is the regularization intensity coefficient.

[0152] S405. Model Validation and Hyperparameter Tuning: During training, the model's performance is evaluated using a validation set, and hyperparameters (such as learning rate, marginal values, and regularization coefficients) are adjusted to optimize model performance. Performance evaluation metrics: Accuracy: The proportion of correctly classified samples. Recall: The model's ability to identify the correct class. F1 Score: The harmonic mean of accuracy and recall.

[0153] By employing methods such as cross-validation, the generalization ability of the model is ensured, avoiding overfitting or underfitting on the validation set. Through this series of steps, the deep learning model can not only accurately identify existing categories but also effectively generalize to new package states not present in the training set, greatly improving the intelligence level and adaptability of the entire express packaging information monitoring system.

[0154] S5. Status Monitoring and Early Warning: The system uses a trained model to monitor the status of express parcels in real time. If the model predicts a potential problem, it will immediately issue an early warning, such as warning that the parcel may have been subjected to vibration or excessive temperature.

[0155] Once a potential anomaly is detected, the system must quickly take the following steps to ensure package security: Warning Generation: Based on the anomaly detection results, generate corresponding warning information, including key data such as the anomaly type, time, and location. For example:

[0156] Vibration warning: If the package is subjected to vibrations exceeding the normal range during transportation, it may cause damage to the contents.

[0157] Temperature warning: The ambient temperature of the package exceeds the safe range, which may affect the quality of temperature-sensitive items.

[0158] Early warning notifications: Early warning information is sent to relevant personnel in real time through multiple channels (such as SMS, email, mobile application notifications, etc.) to ensure timely response. For example:

[0159] Automated response measures: In certain scenarios, the system can automatically trigger response measures, such as adjusting transportation routes, adjusting warehousing environment parameters, or activating backup equipment, to minimize the impact of anomalies on the status of packages.

[0160] Visualization and User Interface: To facilitate user monitoring and management, the system provides an intuitive visual interface that displays real-time monitoring data, prediction results, and early warning information.

[0161] User interaction: Users can view detailed package status information, adjust warning thresholds, configure notification methods, and provide feedback on warning events through the interface to continuously optimize system performance.

[0162] Edge computing deployment: Distributing some computing tasks to edge devices reduces data transmission latency and improves real-time response capabilities.

[0163] Distributed computing frameworks (such as Apache Kafka and Apache Spark) are used for data stream processing and model inference to ensure the system operates efficiently even under large-scale data streams. Model updates and iterations:

[0164] Multi-level risk assessment: Introducing a multi-level risk assessment model to classify and process anomalies of different types and degrees, accurately identifying potential risks:

[0165]

[0166] Where K is the number of risk categories, m k w represents the number of risk indicators of type k. k,i and Sk,i (t) represents the weight and score, respectively.

[0167] A real-time, efficient express parcel status monitoring and early warning system was built by utilizing a trained zero-shot deep learning model. This process encompasses the entire workflow from real-time data acquisition, preprocessing, model inference to anomaly detection and early warning. Key mathematical models and algorithms, such as compatibility functions, risk scoring calculations, and dynamic threshold adjustments, ensure the system's stability and accuracy under various complex environments. Through timely responses from the early warning mechanism, the system effectively safeguards express parcels and improves the intelligence level of logistics management. Combined with a visual interface and continuous optimization strategies, comprehensive monitoring and intelligent early warning of express parcel status are achieved, providing solid technical support for the entire information-based monitoring system.

[0168] Example 2:

[0169] An Internet of Things (IoT)-based information monitoring system for express packaging, applicable to the methods described above, includes sensors and miniature cameras installed on the express packaging for collecting package information and environmental data; a data collection device with edge computing capabilities for real-time data collection and preliminary processing; a zero-shot learning model for learning and generalizing from seemingly unrelated datasets based on patterns learned from training samples; a deep learning model training device for training the collected data using a zero-shot learning algorithm; and an early warning device for real-time monitoring of the express package status using the trained model, providing a warning when the model predicts a problem.

[0170] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0171] It should be understood that the above detailed description of the technical solutions of the present invention with reference to preferred embodiments is illustrative and not restrictive. Those skilled in the art can modify the technical solutions described in the embodiments or make equivalent substitutions for some of the technical features based on reading this specification; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for information monitoring of express packaging based on the Internet of Things, characterized in that: The method includes: Different sensors and miniature cameras are installed on express delivery packaging to collect package information and environmental data; Edge computing devices collect and perform preliminary data collection and processing in real time nearby, improving data processing speed; Building a deep learning model: Building a zero-shot learning model; Training the deep learning model: The zero-shot learning algorithm is used to train the model on the collected data, so that the model can predict new situations that have not appeared in the original training samples; Status monitoring and early warning: The trained model monitors the status of express parcels in real time. If the model predicts a problem, it will immediately issue an early warning. The zero-shot learning model includes the following components: Semantic embedding space: embeds semantic information of categories or states into a high-dimensional vector space, using word vector models; Visual embedding space: The fused feature vector z is projected onto the semantic embedding space to enable cross-space matching. v=W v z+b v Among them, W v and b v These are the weight matrix and bias vector for the linear transformation, respectively, and v is the visual embedding vector; Compatibility function: Define a compatibility function F(v, a) c This is used to measure the similarity between visual embeddings and semantic embeddings; the compatibility function is as follows: Where W is the weight matrix learned, used to capture the relationship between the two; Loss function design: Contrastive loss is used. Where γ is the marginal hyperparameter, c′ is the negative sample class, and the score of the correct class c is at least γ higher than the score of the negative sample c′. The training of the deep learning model includes the following model training process: Forward propagation: The fused feature vector z of each sample in the training set is input into the visual embedding network to generate the visual embedding vector v. Semantic embedding: using a predefined semantic embedding vector a c Represent each category; Compatibility score: Calculate the compatibility score F(v,a) for each pair of visual embeddings and semantic embeddings. c ); Loss Calculation: Based on the marginal contrast loss function, calculate the loss for each sample and accumulate the total loss for the entire training set. Backpropagation and parameter update: The backpropagation algorithm is used to calculate the gradient of the loss function with respect to the model parameters, and the parameters θ are updated through an optimization algorithm. Where η is the learning rate. This represents the gradient of the loss function with respect to the parameters; Regularization and overfitting prevention: Introducing an L2 regularization term to prevent the model from overfitting the training data; The aforementioned status monitoring and early warning system incorporates a multi-level risk assessment model to classify and process anomalies of different types and degrees, accurately identifying potential risks. in, K m represents the number of risk categories. k w represents the number of risk indicators of type k. k,i and S k,i (t) represents the weight and score, respectively.

2. The method for information monitoring of express packaging based on the Internet of Things according to claim 1, characterized in that: The environmental data includes location, temperature, humidity, vibration, and light intensity.

3. The method for information monitoring of express packaging based on the Internet of Things according to claim 1, characterized in that: The data collection and preliminary processing include: data filtering and noise reduction, location data analysis, preliminary image analysis, and vibration data analysis.

4. An information monitoring system for express packaging based on the Internet of Things, wherein the system is applicable to the method as described in any one of claims 1-3, characterized in that: The system includes sensors and miniature cameras installed on the express packaging to collect package information and environmental data; and a data collection device with edge computing capabilities for real-time data collection and preliminary processing. A zero-shot learning model is used to learn and generalize from patterns learned from training samples to seemingly unrelated datasets. The deep learning model training device uses a zero-shot learning algorithm to train the collected data; The early warning device is used to monitor the status of express parcels in real time using a trained model. When the model predicts a problem, the early warning device will issue a warning in real time.

Images